////////////////////////////////////////////////////////////////////////////////
//! Generate the histogram on Single Threaded CPU and Then PTHREADS and Check for Correctness
///////////////////////////////////////////////////////////my_args/////////////////////
void run_test(int num_elements, int num_threads)
{
float diff;
int i;
int *reference_histogram = (int *)malloc(sizeof(int) * HISTOGRAM_SIZE); // Space to store histogram generated by the CPU
int *histogram_using_pthreads = (int *)malloc(sizeof(int) * HISTOGRAM_SIZE); // Space to store histogram generated by the GPU
// Allocate memory for the input data
int size = sizeof(int) * num_elements;
int *input_data = (int *)malloc(size);
// Randomly generate input data. Initialize the input data to be integer values between 0 and (HISTOGRAM_SIZE - 1)
srand(time(NULL)); // add this for real randomness
for(i = 0; i < num_elements; i++)
input_data[i] = floorf((HISTOGRAM_SIZE - 1) * (rand()/(float)RAND_MAX));
printf("Creating the reference histogram. \n");
// Compute the reference solution on the CPU
struct timeval start, stop;
gettimeofday(&start, NULL);
compute_gold(input_data, reference_histogram, num_elements, HISTOGRAM_SIZE);
gettimeofday(&stop, NULL);
printf("CPU run time = %0.10f s. \n", (float)(stop.tv_sec - start.tv_sec + (stop.tv_usec - start.tv_usec)/(float)1000000));
// check_histogram(reference_histogram, num_elements, HISTOGRAM_SIZE);
// Compute the histogram using pthreads. The result histogram should be stored on the histogram_using_pthreads array
printf("\n");
printf("Creating histogram using pthreads. \n");
struct timeval pstart, pstop;
gettimeofday(&pstart, NULL);
compute_using_pthreads(input_data, histogram_using_pthreads, num_elements, num_threads, HISTOGRAM_SIZE);
gettimeofday(&pstop, NULL);
printf("Pthreads run time = %0.10f s. \n", (float)(pstop.tv_sec - pstart.tv_sec + (pstop.tv_usec - pstart.tv_usec)/(float)1000000));
// check_histogram(histogram_using_pthreads, num_elements, HISTOGRAM_SIZE);
// Compute the differences between the reference and pthread results
diff = 0.0;
for(i = 0; i < HISTOGRAM_SIZE; i++)
diff = diff + abs(reference_histogram[i] - histogram_using_pthreads[i]);
printf("Difference between the reference and pthread results: %f. \n", diff);
// cleanup memory
free(input_data);
free(reference_histogram);
free(histogram_using_pthreads);
pthread_mutex_destroy(&mutex);
pthread_exit(NULL);
}
int main(void)
{
int n = NUM_TRAPEZOIDS;
float a = LEFT_ENDPOINT;
float b = RIGHT_ENDPOINT;
float h = (b-a)/(float)n; // Height of each trapezoid
printf("The height of the trapezoid is %f \n", h);
double reference = compute_gold(a, b, n, h);
printf("Reference solution computed on the CPU = %f \n", reference);
/* Write this function to complete the trapezoidal on the GPU. */
double pthread_result = compute_using_pthreads(a, b, n, h);
printf("Solution computed using pthreads = %f \n", pthread_result);
}
void
run_test(int num_elements)
{
float diff;
int i;
/* Allocate memory for the histrogram structures. */
int *reference_histogram = (int *)malloc(sizeof(int) * HISTOGRAM_SIZE);
int *histogram_using_pthreads = (int *)malloc(sizeof(int) * HISTOGRAM_SIZE);
/* Generate input data---integer values between 0 and (HISTOGRAM_SIZE - 1). */
int size = sizeof(int) * num_elements;
int *input_data = (int *)malloc(size);
for(i = 0; i < num_elements; i++)
input_data[i] = floorf((HISTOGRAM_SIZE - 1) * (rand()/(float)RAND_MAX));
/* Compute the reference solution on the CPU. */
printf("Creating the reference histogram. \n");
struct timeval start, stop;
gettimeofday(&start, NULL);
compute_gold(input_data, reference_histogram, num_elements, HISTOGRAM_SIZE);
gettimeofday(&stop, NULL);
printf("CPU run time = %0.2f s. \n", (float)(stop.tv_sec - start.tv_sec + (stop.tv_usec - start.tv_usec)/(float)1000000));
check_histogram(reference_histogram, num_elements, HISTOGRAM_SIZE);
/* Compute the histogram using pthreads. The result histogram should be stored in the
* histogram_using_pthreads array. */
printf("Creating histogram using pthreads. \n");
compute_using_pthreads(input_data, histogram_using_pthreads, num_elements, HISTOGRAM_SIZE);
/* check_histogram(histogram_using_pthreads, num_elements, HISTOGRAM_SIZE); */
/* Compute the differences between the reference and pthread results. */
diff = 0.0;
for(i = 0; i < HISTOGRAM_SIZE; i++)
diff = diff + abs(reference_histogram[i] - histogram_using_pthreads[i]);
printf("Difference between the reference and pthread results: %f. \n", diff);
/* cleanup memory. */
free(input_data);
free(reference_histogram);
free(histogram_using_pthreads);
pthread_exit(NULL);
}
int main(int argc, char **argv)
{
if(argc != 2){
printf("Usage: vector_dot_product <num elements> \n");
exit(1);
}
int num_elements = atoi(argv[1]); // Obtain the size of the vector
/* Create the vectors A and B and fill them with random numbers between [-.5, .5]. */
float *vector_a = (float *)malloc(sizeof(float) * num_elements);
float *vector_b = (float *)malloc(sizeof(float) * num_elements);
srand(time(NULL)); // Seed the random number generator
for(int i = 0; i < num_elements; i++){
vector_a[i] = ((float)rand()/(float)RAND_MAX) - 0.5;
vector_b[i] = ((float)rand()/(float)RAND_MAX) - 0.5;
}
/* Compute the dot product using the reference, single-threaded solution. */
struct timeval start, stop;
gettimeofday(&start, NULL);
float reference = compute_gold(vector_a, vector_b, num_elements);
gettimeofday(&stop, NULL);
printf("Reference solution = %f. \n", reference);
printf("Execution time = %fs. \n", (float)(stop.tv_sec - start.tv_sec + (stop.tv_usec - start.tv_usec)/(float)1000000));
printf("\n");
/* Compute the dot product using the multi-threaded version. */
gettimeofday(&start, NULL);
float result = compute_using_pthreads(vector_a, vector_b, num_elements);
gettimeofday(&stop, NULL);
printf("Pthread solution = %f. \n", result);
printf("Execution time = %fs. \n", (float)(stop.tv_sec - start.tv_sec + (stop.tv_usec - start.tv_usec)/(float)1000000));
printf("\n");
/* Free memory here. */
free((void *)vector_a);
free((void *)vector_b);
pthread_exit(NULL);
}
